METHOD AND DEVICE FOR ERROR MODELING PREDICTIVE CONTEXT ORIENTED TOWARD REVERSIBLE IMAGE WATERMARK

摘要

A method and device for error modeling a predictive context oriented toward a reversible image watermark: constructing a predictor based on an omnidirectional context, then performing adaptive error modeling on the predictive context, obtaining an adaptive error model, and finally feeding back output data of the adaptive error model to the predictor, thereby updating and correcting the predictive context, and correcting a prediction value of the current pixel x[i, j]. The non-linear relationship between the current pixel and the predictive context thereof may be found by means of performing error modeling on the predictive context of the predictor, which is to say that non-linear correlation redundancy between pixels may be found so that the non-linear correlation redundancy between pixels may be effectively removed, which may thus increase embeddable watermark capacity.